Paul J StevensMar 4
Cassandrich

RE: https://chaos.social/@Foxboron/116170859737134271

The blatent use of "AI" to strip copyleft and claim that obvious derivatives are "reimplementations" has begun. This must be fought. Start filing license violation bug reports and abuse reports on projects like this.

(Link here because the quoted post author or their instance seems to have something wrong with their quoting permissions: https://hachyderm.io/@[email protected]ocial/116170859816301768)

Mar 4 at 3:41pmWeb
Show threadRevK Mar 4
@dalias but doing so surely losses all copyright so licence does not matter?
Show threadCassandrichMar 4
@revk No, it doesn't. That's what they're trying to claim and warp the recent court case to say (a case that wasn't even about LLM output) and it's obvious nonsense.
Show threadCassandrichMar 4
@revk Copyright is complex because there are potentially multiple copyright holders when derivative works are involved, and all of their permissions are needed. The court case found that purely machine-produced novelty (which wasn't even LLM and didn't involve derivative works) does not get copyright protection awarded to it. But that has nothing to do with the existing copyrights on any works the output was derived from.
Show threadAlexia Mar 4
@dalias @revk what about the official statement by US Congress? www.congress.gov/crs-product/LSB10922

is that... the "they" you're talking about?
Generative Artificial Intelligence and Copyright Law

Show threadCassandrichMar 4

@alexia @revk No, "they" is the "AI" industry and to some extent software industry that wants to seize and enclose FOSS.

The LoC page you linked agrees largely with what I said except not as strongly; it acknowledges that outputs and models may be derivative and may be infringing but allows that there might be cases that are "fair use".

Show threadAlexia Mar 4
@dalias @revk ty! just wanted to clarify

Though, I think the original reply's point is moreso about how to interpret the human-authorship requirement

if I'm reading it right it'd be evaluated on a pretty much case-by-case basis whether there was enough human authorship but I am not a lawyer by any means
Show threadmaxine 🇵🇸Mar 4
@dalias it’s like the arguments for why treating LLM code and its proponents as actively malicious are all coming from the LLM camps actions! Really hard to claim this was done with anything but active hate of free software as the motivator, since there’s no valid reason to want MIT over LGPL…
Show threadCassandrichMar 4
@maxine There are valid reasons to want MIT over LGPL, but there are not valid reasons to try to forcibly relicense other people's contributions in a way that gives you more power without their consent.
Show threadmaxine 🇵🇸Mar 4
@dalias I’m broadly anti anti-viral licensing and pro-corporate licensing, which is largely how I see the MIT/BSD style licensing side of things, personally, but I think that comes from a great disappointment at free software not actively styling itself as a pro-labour movement reclaiming tools from capital.
Show threadThe Silicon SorceressMar 4
@dalias Oh that's just rotten antisocial behaviour, and the sort of thing that absolutely destroys trust
Show threadUrzlMar 4
@dalias If LLM output can't be copyrighted as SCOTUS seems to have decided, then it also can't be MIT licensed. You can't license what you don't own and you can't own LLM output, it's public domain.
Show threadCassandrichMar 4
@gooba42 And that's ignoring all the fun of what interpretations other jurisdictions might have. Jurisdictions in which things are using that module and now have to block the new version that has no clear entitlement for them to use it. 🤡
Show threadUrzlMar 4
@dalias I'm sure trade secrets or other confidentiality limitations apply to stuff I do at work but the copyright questions are still going to spin off some interesting litigation.
Show threadJoe BrockmeierMar 4

@dalias It's not entirely clear that's what happened here. The author of the new thing seems to be the primary author of the old thing, and I'm not sure that any of the contributors are offended by the license change.

It'll be interesting to see what happens when - not if, when - some company throws something GPL'ed through the LLM license-eradicating machine and tries to pawn that off as something differently licensed. But I'm not convinced that the license change was the purpose of the exercise here.

Show threadThommyMar 4
@dalias imagine if someone working for a company did this but to proprietary code. My worry from the jump has been that it becomes a murky area where we have expansive definition of fair use for me (big corp.), narrow definition for thee (oss dev).
Show threadElladanMar 4

@thomasjwebb @dalias I mean, the Windows source code is available.

What's to stop me from strapping on a joker mask, backing up a sewage tanker in front of Copilot's maw and cackling "exactly this but GPL!" before opening the valves and sprinting away?

What I'm saying is, a world where there's one rule for commercial code and another one for you is exactly what's going to happen.

Show threadnjsgMar 5

@dalias I guess there'll also be a boom of Gen-AI assisted copyfraud - it was already bad to see e.g. unscrupulous publishers slapping new covers on facsimiles and claiming rights on public domain books...

Meanwhile, I suppose Sony BMG is very interested in this application of GenAI - now they can try to cover their tracks better next time they ship a GPL-violating rootkit...